Avalanche transmission and critical behavior in load bearing hierarchical networks

The strength and stability properties of hierarchical load bearing networks and their strengthened variants have been discussed in recent work. Here, we study the avalanche time distributions on these load bearing networks. The avalanche time distributions of the V- lattice, a unique realization of the networks, show power-law behavior when tested with certain fractions of its trunk weights. All other avalanche distributions show Gaussian peaked behavior. Thus the V- lattice is the critical case of the network. We discuss the implications of this result.

💡 Research Summary

The paper investigates avalanche propagation on hierarchical load‑bearing networks, focusing on the statistical distribution of avalanche times (the number of layers traversed by a weight packet). The authors consider a two‑dimensional hierarchical lattice (the “original lattice”) in which each site in layer M is randomly connected to one of its two possible neighbors in layer M + 1 with probability p = ½. The capacity of a site is defined recursively: a site that has no incoming connection can support one unit of weight; otherwise its capacity equals the sum of the capacities of the upstream sites plus one (Eq. 1). This construction is a generalisation of the Coppersmith‑Liu‑Majumdar‑Narayan‑Witten model for granular media and of Scheidegger’s river model.

Within any realisation of the original lattice many clusters appear; the largest is called the maximal cluster, and the path of sites with the greatest cumulative capacity inside this cluster is the trunk. The authors also study a special realisation, the V‑lattice, in which a V‑shaped cluster spans the whole system. One arm of the V is the trunk, the other arm consists of connections parallel to the trunk. This configuration maximises the trunk capacity (denoted W_T) for a given lattice size M.

Avalanche dynamics are defined as follows. A weight W is placed on a randomly chosen site in the first layer. Each site absorbs an amount equal to its capacity W_c and forwards the excess W − W_c to the site it is connected to in the next layer. The process proceeds downward until the excess becomes non‑positive (successful transmission) or the weight reaches the bottom layer. If excess remains after the bottom layer, it is fed back to the first layer and a new downward cycle begins. The avalanche ends either when the excess is exhausted or when a site cannot forward the excess because the downstream site has already saturated its capacity. The avalanche time t is the total number of layers traversed across all cycles.

For the original lattice the authors first examine avalanches with weight equal to the trunk capacity W_T. The resulting distribution P(t) shows no events for t/M < 1 (no path can support more than the trunk) and exhibits a few distinct peaks corresponding to one, two, or three full traversals of the lattice. When the distributions for different lattice sizes (M = 100, 300) are scaled by M, they collapse onto a single curve, indicating a size‑independent shape. When the test weight is reduced to fractions of W_T (0.1 W_T, 0.2 W_T), only a single traversal occurs and the distribution is well described by a Gaussian (Eq. 2) with small standard deviations (σ≈4–6). As the weight fraction increases (0.3 W_T – 0.9 W_T), additional cycles appear, leading to a superposition of Gaussian peaks; at 0.9 W_T the first Gaussian disappears and a third peak emerges, signalling increasingly complex avalanche dynamics.

In stark contrast, the V‑lattice displays power‑law avalanche‑time distributions for weights that are modest fractions of its trunk capacity. For 0.1 W_T and 0.2 W_T the authors find P(t) ∝ t^–α with exponents α≈2.45 and α≈2.96 respectively (Fig. 5). The power‑law regime shrinks as the test weight approaches W_T, and at W_T itself the distribution becomes erratic with no universal form. Moreover, when avalanche times are measured specifically along the trunk of the V‑lattice, a separate scaling law emerges: t ∝ W_f^β with β≈0.33, where W_f is the fraction of the trunk capacity used (Fig. 6). These observations demonstrate that the V‑lattice occupies a critical point of the broader family of hierarchical networks: its asymmetric V‑shaped cluster allows successful transmissions at any layer, generating scale‑free avalanche statistics.

The authors discuss the broader relevance of these findings. Similar avalanche‑type threshold phenomena appear in respiratory airflow networks, voter models, power‑grid cascades, river basins, and directed percolation. The V‑lattice’s ability to support cascades of all sizes suggests that a V‑shaped backbone with a strong trunk could be a design principle for optimising resilience and controllability in real‑world hierarchical systems. The paper concludes by proposing future work on exploiting this critical configuration for network optimisation and on extending the analysis to other dynamical processes.

In summary, the study provides a clear demonstration that a particular structural realisation (the V‑lattice) of a hierarchical load‑bearing network exhibits critical avalanche dynamics characterised by power‑law time distributions, whereas generic realisations produce Gaussian‑like, size‑scaled behaviour. This distinction underscores the importance of network topology in governing cascade phenomena across a wide range of complex systems.